TensorFlow Attention OCR

A TensorFlow model for image text extraction problems.

Forked from https://github.com/tensorflow/models/tree/master/research/attention_ocr

Requirements

1. Install the TensorFlow library.

2. Data. For example,

   datasets
   +-- data
   |   +-- fsns
   |       +-- charset_size=134.txt
   |       +-- fsns-00000-of-00001
   +-- fsns.py
   

Train

To train from scratch:

python train.py \
    --train_log_dir=train_logs

To train a model using pre-trained Inception weights as initialization:

python train.py \
    --checkpoint_inception=inception_v3.ckpt

To fine tune the Attention OCR model using a checkpoint:

python train.py \
    --checkpoint=model.ckpt-399731

Evaluation

python eval.py \
    --train_log_dir=train_logs \
    --eval_log_dir=eval_logs \
    --num_batches=1

Inference

python infer.py \
    --checkpoint=model.ckpt-399731 \
    --image_path_pattern=datasets/data/fsns/testdata/fsns_train_%02d.png